Jupyter MCP Server
An MCP server that lets an AI agent connect to and manage Jupyter notebooks in real time, running against a local Jupyter install, JupyterHub, or a scalable hosted code sandbox from local to the cloud.
Jupyter MCP Server is an MCP server that connects an AI agent to Jupyter notebooks for real-time management and execution, and can scale a code sandbox from a local Jupyter install to hosted services such as Datalayer, Kaggle, Google Colab, Modal, Daytona, E2B, CoreWeave, or Cloudflare. It is free and open source under a BSD 3-Clause license and can point at any Jupyter instance a user already runs, local or JupyterHub, with no account required.
A hosted variant run by Datalayer removes the need to run the server yourself: signing in from a browser approves what an agent may do, work keeps running on the server after the agent disconnects, and Claude Code can connect with one command through a bundled plugin exposing slash commands for opening a notebook, running it, and checking status.
Access is authenticated with an OAuth-issued token scoped to approved operations, such as reading or writing notebooks or executing code, layered under the user's own Datalayer permissions for which notebooks it may reach; personal access tokens remain a simpler option for a CLI or script.